發(fā)布時間：2018-07-07 09:30

  本文選題：擁塞避免 + 自組織社會網(wǎng)絡　； 參考：《大連理工大學》2016年博士論文

【摘要】：傳統(tǒng)的自組織網(wǎng)絡在基礎設施網(wǎng)絡無法搭建的場景(如戰(zhàn)場、自然災害等惡劣環(huán)境)下可為用戶提供快速的通信訪問。自組織社會網(wǎng)絡是傳統(tǒng)自組織網(wǎng)絡的一個分支,該網(wǎng)絡在基礎設施無關(guān)的分布式環(huán)境下,利用用戶的社會屬性,如社會圖、社群、中心度、相似度、聯(lián)系強度和人類移動模式等,為用戶間提供更簡單的通信方式,同時促進資源的高效利用。自組織社會網(wǎng)絡在通信過程中通常交換兩種類型的數(shù)據(jù),分別為社會元數(shù)據(jù)和應用數(shù)據(jù)。其中,前者是為了獲取節(jié)點間的相似興趣或者分享節(jié)點的社會流行度,后者是為了在節(jié)點問傳輸文件或數(shù)據(jù)。但是,由于自組織社會網(wǎng)絡處于無線通信環(huán)境下,其帶寬十分有限,經(jīng)常導致社會元數(shù)據(jù)和應用數(shù)據(jù)的傳輸擁塞。并且,當用戶在單點運行多個社會應用時,很多情況下也會導致網(wǎng)絡擁塞,具體包括：(ⅰ)源節(jié)點和中間節(jié)點的興趣不匹配；(ⅱ)中間節(jié)點隊列容量有限；(ⅲ)數(shù)據(jù)包調(diào)度不當；(ⅳ)目標節(jié)點鏈路容量不足；(ⅴ)不必要的確認消息等。通過研究移動設備用戶間的社會聯(lián)系,自組織社會網(wǎng)絡已經(jīng)成為移動網(wǎng)絡的最重要的形式之一。由于社會網(wǎng)絡用戶的急劇增長,以及多個社會應用的同時運行,自組織社會網(wǎng)絡的擁塞現(xiàn)象十分嚴重。因此,如何在資源稀缺的條件下最大限度的利用網(wǎng)絡帶寬是自組織社會網(wǎng)絡最具挑戰(zhàn)的任務之一。為了解決網(wǎng)絡擁塞問題,傳統(tǒng)的傳輸控制協(xié)議(TCP)采用往返時延機制使用戶共享網(wǎng)絡帶寬。然而,該機制沒有利用節(jié)點的社會屬性,不能有效區(qū)分丟包類型。同時,應用的多樣性和網(wǎng)絡的動態(tài)性也將導致網(wǎng)絡擁塞,進而大大降低中間節(jié)點的服務質(zhì)量。因此,為了避免擁塞,中間節(jié)點經(jīng)常采用棄尾或者隨機早期丟棄技術(shù)來丟棄緩沖隊列中的數(shù)據(jù)包。此外,為了控制中間節(jié)點的擁塞并最先獲取最重要的數(shù)據(jù),節(jié)點通信過程中的數(shù)據(jù)調(diào)度技術(shù)對于自組織社會網(wǎng)絡是十分必要的。通常,自組織網(wǎng)絡采用先進先出機制來進行數(shù)據(jù)包調(diào)度。最后,在多跳組織社會網(wǎng)絡中,可靠的社會聯(lián)系和流行節(jié)點的數(shù)據(jù)傳輸問題在擁塞控制中極其重要。由于數(shù)據(jù)包和確認包使用同一條路徑,使得節(jié)點間競爭無線通信信道,導致數(shù)據(jù)包的碰撞,造成了多跳自組織社會網(wǎng)絡的不可靠性。同時,不必要的確認消息和單節(jié)點的大量數(shù)據(jù)共享都會導致網(wǎng)絡擁塞的發(fā)生。自組織社會網(wǎng)絡中涉及到大量的用戶,每位用戶都希望高效的利用資源,與擁有相同興趣的節(jié)點進行通信。因此,在這樣的網(wǎng)絡中,研究用戶的社交屬性是非常必要的,例如相似性、聯(lián)系強度和社會圖等。而且,為了更好的利用資源,中心度這一社會屬性則是重中之重。所以,如何基于社會感知進行資源共享并且如何通過擁塞控制機制最大限度減少資源浪費是一個非常有意義的研究方向。為了解決自組織社會網(wǎng)絡的資源浪費現(xiàn)象,首先要考慮網(wǎng)絡擁塞問題,使得用戶能在有效時間內(nèi)獲得所需的數(shù)據(jù)包。由于傳統(tǒng)自組織網(wǎng)絡沒有考慮節(jié)點間的社會行為,所以現(xiàn)有的機制并不能滿足自組織社會網(wǎng)絡的需求,而且在擁塞的情況下性能表現(xiàn)不佳。為了更好的控制自組織社會網(wǎng)絡的擁塞問題,本文從網(wǎng)絡屬性(如可用帶寬、節(jié)點隊列容量)和社會屬性(如相似度、接近中心度、度中心度)兩大方面出發(fā),針對數(shù)據(jù)包的控制問題進行深入研究。本文主要解決了如下四個問題：(a)當中間節(jié)點帶寬有限,且只愿意為擁有相同興趣的發(fā)送方分享資源時,如何更好的在發(fā)送方之間分配帶寬?(b)當中間節(jié)點接收到大量發(fā)送方傳輸過來的數(shù)據(jù)包,但自己的隊列容量無法容納所有的數(shù)據(jù)包時,哪種丟包機制和社會屬性對于自組織社會網(wǎng)絡更有意義?(c)中間節(jié)點如何最先傳輸具有社會優(yōu)先權(quán)的數(shù)據(jù)包并保障服務公平性?并且,哪種社會屬性有助于提高節(jié)點間的資源利用率,進而更高效的調(diào)度數(shù)據(jù)包?(d)哪種方法有助于解決擁塞控制機制中的可靠性問題?同時,當數(shù)據(jù)包經(jīng)過多跳節(jié)點進行傳輸時,會遇到由于隱藏/暴露節(jié)點的問題和不必要的確認消息而造成目標節(jié)點和中間節(jié)點容量有限的困難,該方法如何保證資源的充分利用?首先,本文提出了一種基于社會感知的擁塞避免協(xié)議,稱之為TIBIAS。該協(xié)議利用中間節(jié)點的相似度匹配,來提高自組織社會網(wǎng)絡中的資源利用率,進而提升了整個網(wǎng)絡的數(shù)據(jù)傳輸性能。在帶寬資源分配的過程中,TIBIAS將高優(yōu)先權(quán)賦予興趣最相似的TCP連接,并且下層協(xié)議和接收節(jié)點都不需要做任何修改,保證了協(xié)議的獨立性。實驗結(jié)果表明,在鏈路利用率、擁塞窗口不必要的減小、吞吐量和重傳率方面,TIBIAS協(xié)議比現(xiàn)有的擁塞控制協(xié)議表現(xiàn)的更好。其次,本文提出了一種基于生物啟發(fā)的丟包算法(BPD)。該算法模擬免疫系統(tǒng)中受體和表位間的匹配過程來檢測擁塞,基于節(jié)點的流行程度設置丟包概率。受到免疫系統(tǒng)中B細胞刺激過程的啟發(fā),BPD利用節(jié)點的兩種社會屬性,即相似度和接近中心度,來選擇最優(yōu)先的節(jié)點,并為數(shù)據(jù)包公平的賦權(quán)。大量的實驗證明,在平均有效吞吐量、平均丟包率、總吞吐量、時延、獲得帶寬率和超載率方面,BPD算法勝過現(xiàn)有的方法。然后,為了解決自組織社會網(wǎng)絡的擁塞問題,本文提出一種基于用戶流行度的數(shù)據(jù)包調(diào)度算法,稱為Pop-aware。該算法先計算中間節(jié)點的通信負載,接下來使用度中心度這一社會屬性來計算發(fā)送方的社會流行度,進而對輸入流進行優(yōu)先權(quán)值排序。同時,Pop-aware通過計算主動服務率來保證數(shù)據(jù)流所獲得服務的公平性。一系列仿真實驗表明,在控制開銷、總開銷、平均吞吐量、丟包率、包交付率和平均時延方面,該算法性能優(yōu)于現(xiàn)存的調(diào)度算法。最后,本文提出一種適用于自組織社會網(wǎng)絡中流行數(shù)據(jù)的可靠傳輸控制協(xié)議,名為RTPS。該協(xié)議利用度中心度這一社會屬性來計算發(fā)送節(jié)點的社會流行度,并以此為依據(jù)來分配帶寬,從而提高了傳輸可靠性。同時,額外的帶寬則分配給那些社會流行度更高的節(jié)點,且它們的確認消息也擁有更高的優(yōu)先級。另外,通過延遲確認消息的傳輸,RTPS進一步減少了網(wǎng)絡信道競爭和數(shù)據(jù)包丟失。RTPS是以接收方的角度工作的,不需要獲取中間結(jié)點的確切信息。經(jīng)過深入研究得出,在不同的跳距離和不同數(shù)量的TCP并發(fā)數(shù)據(jù)流的情況下,RTPS在吞吐量和傳輸延遲方面均具有明顯的性能優(yōu)勢
[Abstract]:The traditional self-organizing network can provide rapid communication access for users under the unbuilt environment of infrastructure networks such as battlefield, natural disasters and so on. Self organizing social network is a branch of the traditional self-organizing network, which takes advantage of the social attributes of users, such as society under the unrelated infrastructure of infrastructure. Graph, community, community, centrality, similarity, connection intensity, and human mobility pattern, to provide a simpler way of communication among users, and to promote the efficient use of resources. In the process of communication, self organized social networks usually exchange two types of data for social meta data and applied data. Social popularity of similar interests or sharing nodes, which is designed to transmit files or data in nodes. However, since the self organized social network is in a wireless communication environment, its bandwidth is very limited and often leads to the transmission of social metadata and application data. And when users run multiple social applications at a single point, many of them are running. Network congestion can also lead to network congestion, including: the interests of the source node and the intermediate nodes are not matched; (II) the capacity of the intermediate nodes is limited; (III) the inadequacy of the packet scheduling; the insufficient link capacity of the target nodes; the unnecessary acknowledgement messages. Network has become one of the most important forms of mobile networks. Because of the rapid growth of social network users and the simultaneous operation of multiple social applications, the congestion of self organized social networks is very serious. Therefore, it is the most challenging to use the network bandwidth to maximize the use of network bandwidth under the condition of scarcity of resources. One of the tasks. In order to solve the network congestion problem, the traditional transmission control protocol (TCP) uses the round trip delay mechanism to make the user share the network bandwidth. However, the mechanism does not make use of the social attributes of the node, and can not effectively distinguish the type of packet loss. At the same time, the diversity of the application and the dynamics of the network will also cause the network congestion, and then greatly reduce the network congestion. In order to avoid congestion, in order to avoid congestion, the intermediate nodes often use abandoned tail or random early discard technology to discarding the data packets in the buffer queue. In addition, in order to control the congestion of the intermediate nodes and first obtain the most important data, the data scheduling technology in the node communication process is for the self-organized social network. It is very necessary. Generally, the self organizing network uses advanced first out mechanism to carry out packet scheduling. Finally, in the multi hop organizational social network, reliable social connections and data transmission problems of popular nodes are very important in congestion control. Because the data packets and confirmation packets use the same path, the nodes compete for wireless communication. Channels, resulting in the collision of packets, cause the unreliability of a multi hop self-organized social network. At the same time, unnecessary acknowledgement of messages and a large amount of data sharing of a single node will lead to network congestion. A large number of users are involved in the self organized social network, and each user wants to use the resources efficiently and have the same interest. In such a network, it is necessary to study the social attributes of the user in such a network, such as similarity, connection intensity and social map. Moreover, in order to make better use of resources, the social attribute of centrality is the most important. So, how to share resources based on social perception and how to control the congestion through congestion control. To minimize the waste of resources is a very meaningful research direction. In order to solve the waste of resources in the self organized social network, the problem of network congestion should be considered first, so that users can obtain the required data packets within the effective time. Because the traditional self-organizing network does not consider the social behavior among the nodes, the existing network has not taken into account the social behavior of the nodes. In order to better control the congestion of self organized social networks, this paper starts with two aspects of network attributes (such as available bandwidth, node queue capacity) and social attributes (such as similarity, proximity to centrality, degree centrality). The following four problems are discussed in this paper: (a) how to better allocate bandwidth between the sender when the intermediate node has limited bandwidth and is willing to share resources only for the sender with the same interest? (b) when the intermediate nodes receive a large number of packets sent by the sender, but themselves When the queue capacity cannot accommodate all packets, which packet loss mechanism and social properties are more meaningful for self organized social networks? (c) how does the intermediate node first transmit data packets with social priority and ensure service fairness? And which social attributes help to improve resource utilization among nodes and thus more efficient Scheduling packet? (d) which helps to solve the reliability problems in congestion control mechanisms? And when packets are transmitted through multiple hop nodes, it is difficult for the target node and the intermediate node capacity to be limited due to the problem of hidden / exposed nodes and unnecessary acknowledgement messages. To make full use of? First, a congestion avoidance protocol based on social perception is proposed, which is called TIBIAS., which uses the similarity matching of the intermediate nodes to improve the resource utilization in the self organized social network, and then improves the data transmission performance of the whole network. In the process of bandwidth allocation, the TIBIAS will have high priority. The TCP connection with the most similar interest is given, and both the lower layer protocol and the receiving node need no modification to ensure the independence of the protocol. The experimental results show that the TIBIAS protocol performs better than the existing congestion control protocol in the link utilization rate, the unnecessary reduction of congestion window, the throughput and retransmission rate. Secondly, this paper proposes A biologically inspired packet loss algorithm (BPD). The algorithm simulates the matching process between the receptor and the epitopes in the immune system to detect the congestion and sets the packet loss probability based on the popularity of the node. Inspired by the B cell stimulus process in the immune system, BPD uses two social attributes of the node, namely similarity and proximity to the centrality, to choose from the immune system. The BPD algorithm is better than the existing methods in terms of average effective throughput, average packet loss rate, total throughput, delay, bandwidth rate and overload rate. Then, in order to solve the congestion problem of the self organized social network, this paper proposes a user popularity based on this method. The packet scheduling algorithm, called Pop-aware., first calculates the communication load of the intermediate nodes, then uses the social attribute of the degree centrality to calculate the social popularity of the sender, and then prioritize the input stream. At the same time, Pop-aware ensures the fairness of the service by calculating the active service rate. A series of simulation experiments show that the performance of the algorithm is superior to the existing scheduling algorithm in the aspects of control overhead, total overhead, average throughput, packet loss rate, packet delivery rate and average delay. Finally, this paper proposes a reliable transmission control protocol which is suitable for popular data in self-organized social networks, named RTPS., which uses the degree centrality of the protocol. The social attribute is used to calculate the social popularity of the sending node, which is based on the allocation of bandwidth, thus improving the transmission reliability. At the same time, the additional bandwidth is allocated to those with higher social popularity, and their acknowledgement messages have higher priority. In addition, the RTPS is further reduced by delay acknowledgement of the transmission of messages. Less network channel competition and packet dropout.RTPS work on the receiver's point of view and do not have to obtain the exact information of the intermediate nodes. After in-depth study, the RTPS has obvious performance advantages in throughput and delay delay in the case of different jumps and different number of TCP concurrent data streams.
【學位授予單位】：大連理工大學
【學位級別】：博士
【學位授予年份】：2016
【分類號】：TN929.5
，

本文編號：2104538